Article ID: | iaor20042770 |
Country: | Netherlands |
Volume: | 145 |
Issue: | 3 |
Start Page Number: | 645 |
End Page Number: | 659 |
Publication Date: | Mar 2003 |
Journal: | European Journal of Operational Research |
Authors: | Stewart Theodor J., Losa Fabio B. |
The variety of approaches and methods within the broad field of multiple criteria decision analysis (MCDA) can represent an important asset in the development of a meta-approach to decision aid. This requires, however, that their complementarities are recognized, understood, and effectively exploited. The present paper addresses this issue by considering three fundamental cornerstones of the outranking approach to MCDA, i.e., constructivism, partially compensatory preference structures and incomparability. The aim is to demonstrate that while these principles may often be perceived as indicating a fundamental distinction between outranking and multiattribute value theory (MAVT), they do apply equally well to the practical applications of MAVT. First, it will be argued that constructivism is as much a fundamental principle of modern MAVT practice, as it is in the outranking approach. In this context, the axiomatic foundations of MAVT should not be considered in their normative or descriptive features, but in their active role in supporting the learning process. Second, the compensatory nature of the value function models can to some extent be modified to consider, where appropriate, limits to full compensation. This may be achieved by the manner in which the single attribute value functions are defined, and/or by explicit consideration of non-compensatory features as separate decision criteria. Finally, non-completeness of the preference structure, i.e., incomparability in the outranking approach, is seen also to be an implicit result of MAVT application, usually identified during the sensitivity analysis phase. The result is a shift in focus, from emphasis on differences in methods to the intrinsic features of the problem, the decision maker and the decision process, and to the manner in which the different approaches may complement each other in order to enhance decision aid. An explanatory example is used to demonstrate practically the main arguments developed in the paper.